Seven Challenges for the Dehumanization Hypothesis.

Propagandists often compare members of stigmatized out-groups to nonhuman entities such as rats, lice, and snakes. Drawing on these horrifying descriptions, the dehumanization hypothesis proposes that out-group members are viewed as less than human and that being viewed as less than human renders them vulnerable to harm. I offer seven challenges to the dehumanization hypothesis. I argue that, even in supposedly prototypical examples of extreme dehumanization, out-group members are not treated like nonhuman entities. Furthermore, although out-group members may be denied some human qualities and states, they are attributed others. I also argue that there is reason to doubt the hypothesized causal connection between being viewed as less than human and being at risk of harm-some nonhuman organisms are treated with great care, and some groups are harmed because of how their uniquely human qualities are perceived. I close by offering an alternative account of why out-group members are sometimes referred to as nonhuman entities.

Between 1941 and 1945, the Nazis murdered up to 6 million European Jews (Kershaw, 2001). In the spring of 1994, Rwandan Hutus murdered close to 1 million Tutsis (Tirrell, 2012). In the early 2000s, American soldiers tortured and murdered Iraqi prisoners of war at Abu Ghraib (Apel, 2005). Victims of these atrocities, and others like them, were often compared to nonhuman entities—to rats, lice, dogs, and cockroaches (Smith, 2011, 2014).

Research on dehumanization seeks to offer a causal explanation for the apparent association between describing victims of atrocities as nonhuman and inflicting harm on them. Research in this field can be broadly summarized by the following interrelated claims: (a) Victims of intergroup harm are perceived as being similar to nonhuman entities, and, as a result, (b) natural inhibitions against causing them harm are eroded, leading, in extreme cases, to genocide and torture. I refer to these two claims collectively as the dehumanization hypothesis.

The dehumanization hypothesis is ambitious in that it seeks to explain a wide range of phenomena by appealing to the same psychological construct. Blatant dehumanization, in which out-group members are removed from the human category, is thought to play a causal role in genocide and torture (Smith, 2011, 2014, 2016; Viki, Osgood, & Phillips, 2013). More subtle forms of dehumanization, in which out-group members are viewed as “less human” than the in-group, are thought to play a causal role in less extreme acts of harm such as withholding help from individuals in need and feeling reduced concern for their well-being (Andrighetto, Baldissarri, Lattanzio, Loughnan, & Volpato, 2014; Cuddy, Rock, & Norton, 2007; Haslam, 2006; Leyens, Demoulin, Vaes, Gaunt, & Paladino, 2007).

Since its conception, the dehumanization hypothesis has gained considerable prominence in social neuroscience (Harris, 2017; Harris & Fiske, 2006, 2009, 2011), social psychology (Haslam, 2006; Leyens et al., 2007), and philosophy (Smith, 2011, 2014, 2016). Indeed, the hypothesis has become so prominent that the idea that certain groups suffer dehumanization has been described as something of a truism (Smith, 2016).

I argue that, although the dehumanization hypothesis is prima facie reasonable and indeed intuitively compelling, it does not withstand scrutiny. I review the major formulations of the dehumanization hypothesis from neuroscience, psychology, and philosophy. With the use of behavioral data, cognitive data, and historical evidence, I suggest that out-group members are not represented in a similar way as nonhuman entities. Rather, they are likely attributed characteristics that are typical of humans but antisocial in nature such as jealousy, spite, and cunning (Appiah, 2008; Bloom, 2017; Lang, 2010; Manne, 2016). Furthermore, rather than protecting out-group members from harm, being viewed as having these distinctly human attributes may put them at greater risk of harm (Bloom, 2017; Gopnik, 2006; Lang, 2010; Manne, 2016, 2018).

Variants of the Dehumanization Hypothesis

The dehumanization hypothesis is not represented by a single theory but rather by a family of theories from neuroscience, psychology, and philosophy (Harris & Fiske, 2006, 2011; Haslam, 2006; Haslam & Loughnan, 2014; Leyens et al., 2007; Smith, 2011, 2016; Vaes, Leyens, Paladino, & Miranda, 2012). These theories share key underlying assumptions but differ in the particular way in which they characterize the construct of dehumanization. That is to say, each theory offers a somewhat different characterization of what is thought to be “missing” when a group is considered less than human. Below I briefly review the most prominent contemporary theories from philosophy (Smith, 2011, 2016), neuroscience (Harris & Fiske, 2006, 2011), and psychology (Haslam, 2006; Leyens et al., 2007).

Smith (2011) offers a characterization of dehumanization that centers on essentialism. According to this theory, humans are those to whom we attribute a “human essence.” Those to whom we attribute a “subhuman essence” are dehumanized. Smith draws most of his evidence from historical documents, including analyses of propaganda, in which target groups are described as dangerous or disgusting animals such as rats and lice. In a more recent article, Smith (2016) argues that dehumanization involves simultaneously categorizing a group as human in appearance but subhuman in essence. According to this view, simultaneously categorizing someone as human and subhuman gives rise to a feeling of “uncanniness” or creepiness (Smith, 2016).

In contrast, Harris and Fiske (2006, 2011) offer a characterization of dehumanization that focuses on mental-state attribution. According to this theory, when a social group is considered to be human, we attribute mental states to them. When we consider a social group to be “less than human,” we either do not attribute mental states to them or attribute fewer mental states to them (Harris and Fiske, 2006, 2011). Harris and Fiske (2006, 2011) argue that groups perceived to be low in warmth and competence, such as people addicted to drugs and the homeless, are particularly likely to be dehumanized. To the extent that these groups are dehumanized, they will elicit disgust. Harris and Fiske (2006) provide neuroscientific data to support their characterization—participants display less activity in brain regions associated with mentalizing, more specifically the medial prefrontal cortex, when viewing images of homeless individuals and people addicted to drugs than when viewing images of their own group. Convergent evidence comes from behavioral work demonstrating that adults use fewer mental-state words when describing the daily activities of individuals from these groups (Harris & Fiske, 2011) and data suggesting that participants have more stringent criteria for perceiving a mind behind an artificial face when that face belongs to an out-group (Hackel, Looser, & Van Bavel, 2014).

Leyens and colleagues (2007) explicitly departed from the tradition of seeking to explain extreme intergroup harm such as genocide and torture, instead focusing on more subtle manifestations of intergroup biases within contemporary Western society. They conducted an informal survey in which they asked participants what attributes they thought of as uniquely human. Focusing on a subset of their participants’ responses that referenced emotions, Leyens and colleagues argued that groups are subtly dehumanized to the extent that they are thought to be lacking in secondary, or complex, emotions such as pride, guilt, and remorse. They found that individuals typically attribute secondary emotions more strongly to their in-group than to out-groups (Demoulin et al., 2004). To distinguish their work from research on more blatant forms of dehumanization, they termed this hypothesized psychological process “infrahumanization.”

Building on the work of Leyens and colleagues, Haslam and colleagues (Haslam, 2006; Haslam, Bain, Douge, Lee, & Bastian, 2005; Haslam & Loughnan, 2014) sought to characterize dehumanization by first understanding folk perceptions of what it means to be human. Haslam et al. (2005) asked participants to complete two tasks: first, rating a list of attributes for the extent to which they applied to humans and not to other species and, second, rating a list of attributes for the extent to which they were characteristic of humans. Haslam inferred on the basis of their answers to these two questions that there were two forms of humanness. Uniquely human attributes were those that participants rated as distinguishing humans from other species and consisted of civility, refinement, moral sensibility, rationality, and maturity. Human-nature attributes were those that participants rated as characteristic of humans and consisted of emotional responsiveness, interpersonal warmth, openness, agency, and emotional depth. Haslam (2006) postulated two corresponding forms of dehumanization: animalistic dehumanization, in which individuals or groups are thought to possess fewer uniquely human attributes, and mechanistic dehumanization, in which individuals or groups are thought to possess fewer human-nature attributes.

The Relationship Between Dehumanization and Intergroup Harm

Much of the interest in dehumanization stems from the claim that it plays a causal role in intergroup harm. According to the dehumanization hypothesis, humans are naturally inclined to empathize with, and care for, each other. When these barriers against harm are removed, or eroded, individuals are at greater risk of discrimination (Harris & Fiske, 2011; Smith, 2011). Smith (2016, p. 46) describes dehumanization as “a psychological lubricant for the machinery of violence” and argues that it plays “a significant role in facilitating and motivating episodes of genocide, war, slavery, and other forms of mass violence” (Smith, 2014, p. 817). This reasoning is echoed by Harris and Fiske (2011), who argue that “dehumanized perception, a failure to spontaneously consider the mind of another person, may be a psychological mechanism facilitating inhumane acts like torture” (p. 175). Likewise, Haslam and Loughnan (2014) argue that “a major impetus for the study of dehumanization is to understand its profoundly negative consequences. Dehumanization of enemies, victims, and colonized peoples has been associated with pogroms, atrocities, and exploitation” (p. 414) Haslam and Loughnan (2014) further elaborate on this hypothesized causal connection by arguing that dehumanization results in reductions in prosocial behavior, the commission of antisocial acts, and the disinhibition of violence.

Evidence consistent with this proposed causal relation has been drawn from analyses of historical documents. Analyses of propaganda and other historical documents show that regimes that commit genocide and other atrocities often compare their victims to nonhuman entities such as rats, lice, and cockroaches. These examples are widely cited in psychological research on dehumanization (e.g., Harris & Fiske, 2011; Haslam & Loughnan, 2014) but have been most comprehensively studied by philosophers (Smith, 2011; Tirrell, 2012). Smith (2011), for example, focuses primarily on examples from Nazi Germany and American slavery. He points to a plethora of cases in which victims of these systems were compared with, or even described as, subhuman creatures. Furthermore, when interviewed after the atrocities they have committed, perpetrators of mass violence periodically report that they did not view their victims as human (Hatzfeld, 2003).

Further evidence consistent with the claim that dehumanization is causally related to harm comes from lab-based research. This research typically shows that measures of dehumanization correlate with a willingness to endorse harm. For example, Kteily and colleagues showed that the extent to which American participants endorsed the claim that Arabs “seem less highly evolved” than do Americans predicted their endorsement of discrimination against Arabs (Kteily, Bruneau, Waytz, & Cotterill, 2015; see also Kteily & Bruneau, 2017). In related work, Goff, Eberhardt, Williams, and Jackson (2008) found that White participants implicitly associate African Americans with apes and that participants who have been primed with ape-related words are more likely to condone police violence against a suspect but only when that suspect is African American. Other research on the attribution of human qualities has shown that participants’ tendency to deny laborers in sweatshops mental experiences such as the capacity to form plans, as well as to feel emotions such as love and pain, correlates with their willingness to endorse the use of sweatshop labor (Rai, Valdesolo, & Graham, 2017).

Seven Challenges for the Dehumanization Hypothesis

Challenge 1: comparisons to nonhuman entities are not reserved for out-groups

A key source of evidence in favor of the dehumanization hypothesis comes from real-world examples of situations in which members of certain out-groups have been compared to nonhuman entities (Smith, 2011, 2014, 2016; Tirrell, 2012). Although intuitively compelling, these examples alone are not sufficient to conclude that out-group members are viewed as more similar to nonhuman entities than are in-group members. If using historical associations between groups and animals as one source of evidence for the dehumanization hypothesis, it is crucial to search for disconfirmatory as well as confirmatory cases (for a review of research on the confirmation bias, see Nickerson, 1998). Surveying real-world examples more broadly, it is clear that comparisons to nonhuman entities are not always used as a way to insult or demean (Haslam, Loughnan, & Sun, 2011). Comparisons to animals can be used to compliment an individual and even to highlight some of their prototypically human virtues. For example, the epithet “Lionheart” was intended to emphasize the bravery of Richard I of England in battle. Although comparing a person to a monkey can sometimes be deeply offensive, using “little monkey” as a term of endearment to describe a toddler might emphasize that they are charming and mischievous. Likewise, comparing an athlete to a machine might emphasize their perseverance and skill. Social groups also often invoke nonhuman entities to refer to themselves. For example, sports teams often have nonhuman entities as their emblems and might refer to themselves as lions, bulls, blades, gunners, or eagles (see Fig. 1).

Fig. 1.

Groups often refer to themselves in terms of nonhuman entities. Examples of this are the Tottenham Hotspur emblem, depicting a cockerel (left); the Chicago Bulls emblem, depicting a bull (center); and the Gadsden flag, depicting a rattlesnake (right). The image in (c) is used under an Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) from Wikimedia Commons user Vikrum.

It might be possible to counter that whereas comparisons to some nonhuman entities are complimentary, comparisons to others are, for some as yet unspecified reason, dehumanizing. However, in-group members sometimes compare themselves to the same supposedly “disgusting and dangerous” animals to which out-groups are often compared. For example, the Tutsis in Rwanda were often compared to snakes by propagandists (Tirrell, 2012), but the American revolutionaries often compared themselves to snakes. The Gadsden flag of the American Revolution depicts the American people as a rattlesnake ready to bite the British Empire (Rankin, 1954; see Fig. 1). Taken together, these examples demonstrate that comparisons to animals are neither, in and of themselves, problematic nor necessarily a reflection of a deep difference in how members of in-groups and out-groups are represented. The challenge for the dehumanization hypothesis is to explain why comparisons to nonhuman entities are sometimes taken as evidence for dehumanization and sometimes not.

Challenge 2: out-group members are often described in ways that apply only to humans

The dehumanization hypothesis proposes that, to the extent they are dehumanized, out-group members are perceived in a similar way as nonhuman entities, most commonly animals or automata (Haslam, 2006; Haslam & Loughnan, 2014; Haslam et al., 2011; Smith, 2011). Although there may be occasions on which out-group members are described in ways that are equivalent to how animals and automata are described, these cases are much less common than they first appear. A careful reading of propaganda and hate speech reveals that target groups are often described in ways that apply only to humans (Bloom, 2017; Manne, 2016). For example, Nazi propaganda often referred to Jewish people as criminals, murderers, enemies, and traitors (Keen, 1992). These terms are readily applied to humans but make little sense if applied to animals or other nonhuman entities (Manne, 2016). A rat or an automaton cannot be a criminal or a traitor. The use of these descriptors thus suggests that the Jewish population was at least implicitly represented as human.

It might be possible to counter this critique by arguing that out-groups are sometimes dehumanized, or dehumanized by some writers, and discriminated against in different ways at other times and by other writers. However, this idea does not fit with the historical evidence either—a variety of metaphors often appear within the same piece of propaganda. For example, in “The Jewish World Plague,” Hermann Esser (1939) describes Jews not only as weeds, parasites, and worms but also as swindlers, thieves, beggars, and deceivers. In what can be translated from the German as “Knowing the Jew Means Understanding the Meaning of the War,” distributed by the Reich Propaganda Office (Reichspropagandaleitung, 1944), Jewish people are described not only as parasites, mistletoe, and an infection but also as “deadly foes” having committed slander and desirous of world domination.

As Manne (2016) and Appiah (2008) have pointed out, even in the act of referring to a group as lice or vermin, propagandists reveal an implicit recognition of the difference between their targets and the animals to which they are compared. There is no sense in consistently reminding a rat that it is, in fact, a rat (Manne, 2016). Rather, the power of the metaphor comes from the recognition that the two entities are different (Bloom, 2017). If victims of intergroup harm are not viewed as less than human even in these supposedly prototypical cases, then the onus is on proponents of the dehumanization hypothesis to explain when out-groups are dehumanized.

Challenge 3: being associated with a nonhuman entity is not equivalent to being seen as similar to that nonhuman entity

Psychologists have typically relied more heavily on evidence from lab-based research than on evidence from historical records. For example, data showing that White Americans implicitly associate African Americans with apes in lab-based tasks is often cited as an example of dehumanization (Goff et al., 2008; Goff, Jackson, Di Leone, Culotta, & DiTomasso, 2014). Although these implicit associations are clearly deeply problematic they are not convincing evidence that African Americans are perceived as less human than are White Americans or as akin to apes. Associative connections between two stimuli do not necessarily suggest that they will be viewed as equivalent or as similar to each other. To borrow an example from lab-based research on animal learning, a rat may come to associate a gray square with food through repeated presentations of the square in conjunction with food (Rescorla & Wagner, 1972). However, the rat does not come to think of the gray square as being food or similar to food (and would not try to eat it). Relatedly, cultural pairings may lead to associations between African Americans and apes, which reveal themselves in nonverbal tasks in the lab but that are not, on their own, evidence that participants in those studies viewed African Americans as similar to apes.

Challenge 4: out-group members may be denied some mental states but they are frequently attributed others

In some theories, dehumanization is characterized as a failure to consider the mind of another person (Harris & Fiske, 2011; see also Hackel et al., 2014; Harris & Fiske, 2006, 2011; Leyens et al., 2007; Rai et al., 2017; Waytz, Gray, Epley, & Wegner, 2010). In the seminal study advocating for this view, Harris and Fiske (2006) provided evidence that, when presented with images of homeless individuals and people addicted to drugs, participants display less activity in brain regions associated with mentalizing.

This characterization of dehumanization does not appear to accurately characterize how out-group members are perceived. There are certainly situations in which the thoughts and emotions of out-group members are not sufficiently salient to dominant majorities (Harris & Fiske, 2011). However, I predict that out-group members are not thought to lack all mental states or even to hold mental states to a lesser extent than in-group members. Although out-group members may be denied some mental states, they are likely attributed others (Hackel et al., 2014). To analyze Harris and Fiske’s own example a little more closely, people addicted to drugs are often criticized because of how observers represent their mental states—they are perceived as greedy and lacking in self-control (Corrigan, Kuwabara, & O’Shaughnessy, 2009). Further evidence that this is the case can be drawn from historical records. Even in supposedly prototypical examples of blatant dehumanization, target groups are described in terms of mental states. Propaganda often references the mental states of its targets to generate hatred against them. In Nazi propaganda, for example, Jewish people were often attacked for their supposed cunning, malice, and scheming against the Nazi regime (Bytwerk, 2004; Keen, 1992; Kershaw, 2001).

Furthermore, it is not clear why possessing certain mental states to a lesser extent than in-group members should be conceptualized as dehumanization. Equally problematic for the dehumanization hypothesis are cases in which in-group members are denied some mental states or afforded them to a lesser extent than out-group members. For example, in-group members are typically judged to be less deceptive, sneaky, and ruthless than are out-group members (Dunham, 2018). We are thus left with a peculiar situation in which biases in mental-state attribution are sometimes characterized as dehumanization and sometimes not.

Challenge 5: out-group members are granted some uniquely human attributes

Haslam and colleagues argue that groups can be dehumanized in one of two respects. They can fall victim to animalistic dehumanization in which they are thought to possess attributes such as civility, refinement, and rationality to a lesser extent than do the in-group. Alternatively, they can fall victim to mechanistic dehumanization, in which they are thought to possess attributes such as emotional responsiveness, interpersonal warmth, and agency to a lesser extent than do the in-group (Haslam, 2006; Haslam et al., 2005; Haslam & Loughnan, 2014).

On closer inspection, Haslam and colleagues’ characterization of dehumanization appears incomplete. In particular, it omits antisocial attributes such as jealously, spite, dishonesty, and disloyalty. These antisocial attributes are not salient in Haslam and colleagues’ characterization of the concept of humanness, and yet they make sense only when applied to humans. It would be extremely unusual, and most likely inappropriate, to describe an animal or a machine as disloyal or spiteful, for example (Manne, 2016, 2018).

I propose that this problem arises because of the way in which Haslam and colleagues sought to identify participants’ lay concept of humanness. Haslam and colleagues assumed that they could characterize the lay concept of humanness by asking participants (a) which from among a list of attributes they felt distinguished humans from other species and (b) which attributes they thought were characteristic of humans. This approach is at odds with decades of research in the cognitive psychology of categorization (Barsalou, 1989; Bellezza, 1984; McNamara & Sternberg, 1983; Medin, 1989). Research in this area demonstrates that the attributes that appear typical of a category will vary depending on the context (Smith & Medin, 1981). Thus, the attributes that appear typical of the category “human” will differ depending on the comparison point. As a thought experiment, imagine that instead of asking his participants what distinguishes humans from other species, Haslam and colleagues had asked their participants what distinguishes humans from another nonhuman category—angels (see Fig. 2). I predict that more antisocial qualities such as greed, jealousy, laziness, and spite would have been listed as typical of humans in this context. Somewhat different attributes again would be salient to participants if the comparison point were zombies, robots, rats, or dolphins. It is no coincidence that Haslam and colleagues asked their participants two questions and found evidence for two characterizations of the concept of humanness. Had they asked their participants a third question, for example, how humans differ from angels, they may well have found evidence for three forms. This leaves us with a puzzle in which being perceived as somewhat lacking civility and warmth is thought to constitute dehumanization, whereas being seen as somewhat lacking in spite and jealousy, also uniquely human attributes, is not thought to constitute dehumanization.

Fig. 2.

The salience of particular attributes depends on the comparison point. Attributes listed when comparing humans to gorillas will be different from those listed when comparing humans to angels. The image on the left is used under an Attribution 2.0 Generic (CC BY 2.0) license from the photographer, Derek Keats. The image on the right is used under an Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license from Wikimedia Commons user Victuallers.

When we consider these uniquely human but antisocial attributes as part of what it means to be human, it is no longer clear that out-groups are perceived as less human than are in-groups. Out-group members are often thought to possess antisocial attributes such as cunning and spite. Indeed, they are thought to possess them to a greater extent than do members of the in-group (Fiske, Cuddy, Glick, & Xu, 2002).

Challenge 6: groups are often persecuted because of their perceived humanity

The dehumanization hypothesis is based on the assumption that being perceived as human tends to offer protection from harm. In this view, humans naturally empathize with and care for each other (Harris, 2017; Harris & Fiske, 2011; Smith, 2011). When a group is perceived to be less human, or less than human, some of these natural responses of care and empathy are thought to be eroded, rendering the group more vulnerable to harm (Haslam, 2006; Haslam & Loughnan, 2014; Harris, 2017; Smith, 2011).

This confidence in humans’ desire to protect and care for each other is partly misplaced (Lang, 2010). Although humans do often care for each other, certain groups may be targeted because they are believed to possess uniquely human mental states and attributes (Bloom, 2017; Manne, 2018). It seems relatively uncontroversial to predict that perceiving a group as having human qualities such as spite and cunning will make them appear to be a threat—a threat that, in extreme cases, will need to be controlled or destroyed (Keen, 1992). Likewise, believing a group exists in human-specific social relations to a dominant majority, such as being enemies, traitors, rapists, or criminals, will likely make them appear morally responsible for their actions (Brown & Webb, 2007). This is one reason why violence against target groups can take on a moral quality (Rai & Fiske, 2014; Rai et al., 2017). Whereas eradicating vermin might be seen as desirable, the lynching or mass murder of humans can be framed as a moral crusade (Dray, 2003; Keen, 1992; Rai & Fiske, 2014). the lynching of African Americans was often presented as exacting justice on the guilty (Dray, 2003). Likewise, propaganda in Nazi Germany often described the German population as the victim of Jewish aggression and plots (Keen, 1992). In lab-based research, Rai and colleagues (2017) showed that reflecting on the mental states of a potential target increases morally motivated violence against them.

Manne (2018) has relatedly pointed out that victims of atrocities are humiliated and tortured because their abusers at least implicitly recognize their humanity. Whereas rats might be killed, they are not forced to simulate sex acts with other rats or to watch their fellow rats die. Nor would there be any sense in publicly humiliating or killing a rat to serve as a warning to other rats. Yet human victims often suffer these indignities (Bloom, 2017; Brown & Webb, 2007; Dray, 2003). Thus, even in these supposedly prototypically dehumanizing actions, perpetrators implicitly acknowledge the difference between their victims and nonhuman entities in the type of harm that they inflict on them (Bloom, 2017; Manne, 2016, 2018).

Challenge 7: being seen as less than human is not necessarily a risk factor for harm

Equally problematic for the dehumanization hypothesis are cases in which individuals are thought to lack characteristically human qualities and yet are not subjected to harm. Consider the example of a baby. As Smith (2011) has observed, babies are not typically considered as having complex thoughts and beliefs. Nor are they thought to possess typically human attributes such as refinement, civility, and maturity. Nevertheless, they are treated with the utmost care and devotion by their parents and caregivers. Indeed, treating young children as more mature and rational than they are would most likely be harmful to them (Goff et al., 2014).

Even when organisms are clearly viewed as outside of, or indeed beneath, the human category they are not necessarily at risk of harm. Although many animals are treated badly by humans, some are treated with care. People donate substantial funds to conservation efforts focused on animals such as the giant panda. Even allegedly dangerous and disgusting animals such as snakes are protected by conservation efforts. Likewise, family pets, although typically thought of as being less important than humans, are usually treasured and protected by their human families (see Fig. 3).

Fig. 3.

Examples of entities viewed as “less than human” that are often treated with care.

In certain circumstances, animals are actually treated with greater care than are humans. In the United Kingdom, the Royal Society for the Prevention of Cruelty to Animals receives substantially more donations than Shelter, the best-known charity for the homeless (Shelter, 2018; YouGov, 2018). On an individual level, Hitler was responsible for the murder of millions of innocent people and yet he adored his pet dogs (Kershaw, 2001). Objects, too, are sometimes treated with great care. The campaign to save the Notre Dame chapel after the 2019 fire raised hundreds of millions of euros (Lyons & Busby, 2019). These examples suggest that it is not being perceived as less than human per se that puts a group at risk of harm.

Is it possible to explain the prevalence of nonhuman metaphors in propaganda without recourse to the concept of dehumanization?

I have outlined seven challenges for the dehumanization hypothesis. I have argued that although there are clearly differences in how members of different social groups are perceived, these differences do not seem best characterized as a psychological process of dehumanization. This leaves us with an open question, however. If out-group members are not perceived as less than human, then how are we to explain the apparent prevalence of comparisons to nonhuman entities in propaganda and other forms of hate speech? Below I sketch an alternative theoretical account of why comparisons to nonhuman entities are sometimes associated with harm.

Propagandists use comparisons to certain nonhuman entities to persuade audiences to view their target group in a negative light. By consistently pairing representations of a target group with negatively valenced stimuli such as disease and lice, propagandists encourage observers to negatively evaluate that group. Note that this mechanism is not specific to references to subhuman entities. Rather, conditioning can take place by pairing the group with any negative concept, including words that evoke human-specific social relations such as enemy, rapist, criminal, and traitor (Siegel & Allan, 1996).

Referring to a target group as a certain type of nonhuman entity also licenses particular inferences about that group and its supposed traits (Tirrell, 2012). For example, by evoking the idea that a target group are like cockroaches, a propagandist might convey the idea that they are disgusting and of low status. By evoking the idea that a target group are like monkeys, a propagandist might convey the idea that they are stupid and unsophisticated. Metaphors of this type can also license particular behaviors (Tirrell, 2012). For example, whereas comparing a target group to oxen might imply that they ought to be enslaved, comparing a target group to cockroaches might imply that they ought to be exterminated. Again, this mechanism is not specific to subhuman categories. Referring to a target group as the enemy licenses the inference that they are a threat and need to be defeated.

To the extent that comparisons to nonhuman entities are more common in propaganda and hate speech than they are in other forms of communication, it may be because they are especially effective ways of conveying multiple negative messages simultaneously. For example, whereas using the insult “enemy” conveys the idea that the target group is a threat, using the insult “louse” not only conveys that the group is a threat but also that the group is contaminating, of low status, and homogeneous in character.

Nonhuman metaphors can also be used to communicate negative messages to the target group themselves—they can be used to humiliate and threaten (Bloom, 2017; Lang, 2010; Manne, 2016). In these cases, it is not the message that the group is less than human per se that is problematic. Rather, these nonhuman comparisons are offensive to the extent that they imply the target group is of low status, disgusting, and threatening. In Western culture, it would be bizarre but not particularly offensive to repeatedly refer to a group as dolphins or pandas (both of which are considered “less than human”), whereas it would be deeply offensive to refer to a group as rats. The insult “rat” is offensive because of learned associations between rats, diseases, contaminations, and threats.

It is worth pointing out that comparisons to nonhuman entities are not always used intentionally to persuade or humiliate. If listeners repeatedly hear a target group compared to negatively valenced animals such as rats or snakes then they will come to associate that group with those particular entities (Devine, 1989). Once this association is in place, when listeners hear the target group mentioned, images of the associated entity will be automatically activated (Goff et al., 2008). Automatic associations between out-groups and animals, or indeed other negatively valenced categories, can leak in to communication outside of conscious awareness. For example, an individual who automatically associates a particular group with apes might use terms such as “jungle,” “wild,” and “creature” when describing that group even when they have no intention of exacerbating animosity toward the group in question.

Note that comparisons to nonhuman entities are not always used to criticize, humiliate, and demean. They can also be invoked to compliment a group and extol their virtuous intentions and attributes. For example, consistently referring to a group as lions might activate associated inferences about courage and dominance. Referring to a group as angels might emphasize the ways in which their mental states and attributes are worthy of respect and admiration. Again, according to my perspective, references to nonhuman entities are not the only way to convey these messages, but they are one effective means by which to do so.

Directions for future research

The seven challenges outlined here bring much of what we thought we knew about dehumanization into question. These challenges also suggest a number of priorities for future research that will enable the field to distinguish between the dehumanization hypothesis and the alternative I have outlined here. First, it will be crucial to conduct systematic content analyses of historical documents to establish the incidence of nonhuman comparisons in propaganda and other forms of hate speech. In conducting these content analyses, it will be of paramount importance to avoid the confirmation bias, whereby historical records are searched only for examples that provide positive evidence for the dehumanization hypothesis (i.e., cases in which target groups are referred to as nonhuman entities). The most problematic cases for the dehumanization hypothesis include passages in which target groups are described in terms of uniquely human attributes, mental states, and emotions.

Another priority for future research is to reassess how the construct of dehumanization is operationalized. Thus far, much psychological research on dehumanization has focused on the attribution of broadly positive and prosocial qualities such as civility, rationality, warmth, pride, and guilt (Haslam, 2006; Leyens et al., 2007). Future research must incorporate insights from context-dependent categorization to understand whether adults also associate antisocial attributes such as jealousy, spite, cunning, greed, and dishonesty with the concept of humanness.

If it transpires that participants do associate antisocial attributes such as these with the concept of humanness, then further empirical studies can determine whether out-group members are thought to possess all uniquely human attributes to a lesser extent than in-group members (as predicted by the dehumanization hypothesis) or only those that are prosocial in nature. Furthermore, future work can reassess the hypothesized causal relationship between the attribution of uniquely human qualities and intergroup harm by evaluating whether describing a potential victim in terms of uniquely human but antisocial attributes places them at greater or lesser risk of harm.

Conclusion

According to the dehumanization hypothesis, when a dominant majority describes an out-group as rats or lice they do so because they believe that group to be less than human. This argument has an elegant simplicity and an intuitive appeal. Not surprisingly, therefore, it has become extremely prominent in social psychology, social neuroscience, and certain areas of philosophy (Harris & Fiske, 2006, 2009; Haslam, 2006; Haslam & Loughnan, 2014; Leyens et al., 2007; Smith, 2011, 2014, 2016). Indeed, it is all but accepted as fact that certain groups suffer dehumanization (Smith, 2016).

I have argued that the explanatory power of this hypothesis is considerably more restricted than it first appears. Although there may be some cases in which out-group members are genuinely believed to be less than human, there is not yet convincing evidence that this is a common phenomenon. Comparisons to nonhuman entities are not restricted to out-groups; nor are out-groups regularly described in similar ways to nonhuman entities. Although members of out-groups are often perceived to be lacking in some human qualities and attributes, they may well be thought to possess other human attributes to a greater extent than do the in-group (Bloom, 2017; Lang, 2010; Manne, 2016). In addition, there are other plausible explanations for why comparisons to nonhuman entities may be common in propaganda and other forms of hate speech. It remains for future empirical research to determine the relative merits of the dehumanization hypothesis and the alternative view I have suggested here.

If supported by future empirical research, the argument I have advanced has important implications for our understanding of how to bring about social change. Inspired by the dehumanization hypothesis, researchers have shown an increasing interest in developing interventions to reduce intergroup harm that focus on “humanizing” out-groups (Albarello & Rubini, 2012; Gaunt, 2009). My critique suggests that other routes to reducing intergroup harm may prove more effective. Encouraging dominant majorities to reflect on the humanity of out-groups could even backfire in certain circumstances, if, for example, it leads to increased focus on the supposedly antisocial mental states and attributes of the group in question.